Programming Research Group Technical Monograph PRG-71

Roberto Garligliano and Derek Long

February 1989, 195 pages, ISBN 0-902928-53-8

We address the problem of how to deal with inaccurate, incomplete and changing information. We proceed by formally comparing existing reasoning systems in order to precisely define which features are needed and which should be avoided. In order to do so, we construct a formal theory, which we call homogeneity theory. The equivalence of the transformation rule of identity, for each reasoning system, to some expressions of homogeneity theory is proved. An order is then induced among the systems analysed using the expressions of homogeneity theory.

Some natural criteria are formally defined in order to evaluate the respective power of the systems, which lead to a second order of the systems, embedded in the first. An intermediate model and finally the model, called adapative reasoning system, are specified. They are proved to be stronger than the systems previously examined, according to the criteria defined.

The central idea in the adaptive reasoning systems is the attempt to recognize patterns of behaviour in sources of information, or areas of interest, and regulate the reliability of the sources and the stability of the areas accordingly. The adaptive reasoning system is equipped with yardsticks against which to judge and regulate its own performance.